R E S E A R CH AR T I C L E

Stable isotope variation in savanna chimpanzees(Pan troglodytes verus) indicate avoidance of energetic challengesthrough dietary compensation at the limits of the range

Erin G. Wessling1,2 | Vicky M. Oelze1,3 | Henk Eshuis1 | Jill D. Pruetz4 | Hjalmar S. Kühl1,2

1Department of Primatology, Max Planck

Institute for Evolutionary Anthropology,

Leipzig, Germany

2German Centre for Integrative Biodiversity

Research (iDiv), Leipzig, Germany

3Department of Anthropology, University of

California Santa Cruz, Santa Cruz, California

4Department of Anthropology, Texas State

University, San Marcos, Texas

Correspondence

Erin G. Wessling, Max Planck Institute for

Evolutionary Anthropology, Deutscher Platz

6, 04103 Leipzig, Germany.

Email: erin_wessling@eva.mpg.de

Funding information

Krekeler Foundation; Iowa State University;

Max Planck Society Innovation Fund; Max-

Planck-Gesellschaft; National Geographic

Society

AbstractObjectives: Food scarcity is proposed to be a limitation to chimpanzees at the limits of their

range; however, such a constraint has never been investigated in this context. We investigated

patterns of δ13C and δ15N variation along a latitudinal gradient at the northwestern West Afri-

can chimpanzee (Pan troglodytes verus) range limit with the expectation that isotope ratios of

chimpanzees at the range limit will indicate different dietary strategies or higher physiological

constraints than chimpanzees further from the edge.

Materials and methods: We measured δ13C and δ15N values in hair (n = 81) and plant food

(n = 342) samples from five chimpanzee communities located along a latitudinal gradient in

Southeastern Senegal.

Results: We found clear grouping patterns in hair δ13C and δ15N in the four southern sites com-

pared to the northernmost site. Environmental baseline samples collected from these sites

revealed overall higher plant δ15N values at the northernmost site, but similar δ13C values across

sites. By accounting for environmental baseline, Δ13C and Δ15N values were clustered for all

five sites relative to total Pan variation, but indicated a 13C-enriched diet at the range limit.

Discussion: Clustering in Δ13C and Δ15N values supports that strategic shifting between

preferred and fallback foods is a likely ubiquitous but necessary strategy employed by these

chimpanzees to cope with their environment, potentially allowing chimpanzees at their limits to

avoid periods of starvation. These results also underline the necessity of accounting for local

isotopic baseline differences during inter-site comparison.

KEYWORDS

carbon, niche, nitrogen, δ13C, δ15N

1 | INTRODUCTION

Carbon and nitrogen isotope values (δ13C and δ15N) are advanta-

geous tools used in disciplines like paleoanthropology and ecology

due to their ability to inform us on the diet and habitat of an indi-

vidual. Fundamental to the use of stable isotopes as an indicator of

diet is the idiom “you are what you eat” (Kohn, 1999), in that an

individual's isotopic composition is a reflection of the isotopic ratio

of the food items it consumes. These isotopes provide semi-quantified

information on the diet of the individuals from which the tissues

are sampled (Schwarcz & Schoeninger, 1991), and can be used, for

example, to differentiate between photosynthetic pathways in

plant-based diets (DeNiro & Epstein, 1978), or used as indicators of

trophic level (for example as between plant- and animal-based

diets: Minagawa & Wada, 1984). As a result, these isotopes have

served as the basis upon which studies of dietary composition,

change, and niche partitioning have been undertaken in a wide range of

taxa (e.g., Balasse, Bocherens, Mariotti, & Ambrose, 2001; Caraveo-

Patino, Hobson, & Soto, 2007; Codron, Lee-Thorp, Sponheimer, de

Ruiter, & Codron, 2006; Codron, Lee-Thorp, Sponheimer, De Ruiter, &

Codron, 2008; Harrison et al., 2007; Koch et al., 1995; Lee, Schell,

McDonald, & Richardson, 2005; Zazzo et al., 2007).

Received: 21 June 2018 Revised: 20 December 2018 Accepted: 2 January 2019

DOI: 10.1002/ajpa.23782

Am J Phys Anthropol. 2019;168:665–675. wileyonlinelibrary.com/journal/ajpa © 2019 Wiley Periodicals, Inc. 665

Because tissues used for stable isotope analyses (e.g., hair, teeth,

and bone) reliably record the nature of isotopic intake, isotope ecology

has recently gained popularity in primatological research (Crowley,

Rasoazanabary, & Godfrey, 2014; Oelze et al., 2016; Sandberg,

Loudon, & Sponheimer, 2012; Schoeninger, 2009; Schoeninger,

Iwaniec, & Nash, 1998). Furthermore, a particular advantage of

using δ13C and δ15N as dietary indicators is the possibility to incor-

porate otherwise unobtainable data on groups unhabituated to

researcher presence in the face of rapid species decline (e.g., Kühl

et al., 2017).

Pan species are frequently used as referential models for hominin

ecology and dietary reconstruction; therefore these species have like-

wise represented a particular focus within primate isotopic research

(e.g., Carter, 2001; Fahy, Boesch, Hublin, & Richards, 2015; Fahy,

Richards, Riedel, Hublin, & Boesch, 2013; Macho, 2016; Macho &

Lee-Thorp, 2014; Oelze et al., 2016; Oelze, Head, Robbins, Richards, &

Boesch, 2014; Schoeninger, Moore, & Sept, 1999; Schoeninger,

Most, Moore, & Somerville, 2016; Smith, Morgan, & Pilbeam, 2010;

Sponheimer et al., 2006; van Casteren et al., 2018). As a result, Pan

isotope ecology has developed into a growing discipline in which both

broad- and fine-scale dietary variation within and between individuals

have been detected noninvasively. For example, Fahy et al. (2013)

demonstrated that interindividual differences in hunting prowess

among wild chimpanzees is detectable using δ15N values on the basis

that better hunters are likely to have a higher proportion of meat in

their diet and therefore have higher δ15N values than less successful

hunters. Additionally, researchers have begun to investigate the use

of nitrogen isotopes for detecting periods of nutritional stress in

multiple taxa (e.g., Kempster et al., 2007; McCue & Pollock, 2008),

more recently extending to great apes (Deschner et al., 2012; Vogel

et al., 2011). Such patterns are built upon the tenet that increases in

the metabolization of animal body tissue (whether extrinsic via meat

consumption or intrinsic via fasting or starvation) effect observable

δ15N increases in the tissues of the consumer (Fuller et al., 2005).

However, in larger-scale comparisons of δ13C and δ15N values

across multiple ecological conditions potential conclusions become

less straightforward. Initial studies attempted to evaluate aspects of

primate diet and ecology using tissue samples alone, yet the recogni-

tion of the variability of tropical ecosystems led researchers to

embrace the incorporation of environmental context into cross-site

comparisons (Crowley et al., 2014; Oelze et al., 2011; Oelze et al.,

2016). For example, one prominent contextual measure of isotope

system variability was mean annual precipitation (MAP). However,

attempts to explain isotopic variation between chimpanzee sites using

MAP have resulted in contradictory findings, as significant empirical

support for this relationship exists for δ13C values, but widespread

consensus is lacking for δ15N (Loudon, Sandberg, Wrangham, Fahey, &

Sponheimer, 2016; Oelze et al., 2016; Schoeninger et al., 2016).

Although attempts at explaining climatic contributions aids in

understanding broad-scale patterns of variation in Pan isotopes, the

relevance of the ecological context in isotopic comparisons across

varied habitats necessitated the establishment of environmental

isotopic baselines before fine-scale variation could be evaluated.

Isotopic variability of Pan food items (Carlson & Crowley, 2016;

Carlson & Kingston, 2014; Loudon et al., 2016) and the incorporation

of these items into body tissues (Tsutaya, Fujimori, Hayashi, Yoneda, &

Miyabe-Nishiwaki, 2017) has recently received greater focus in the

literature. Still, scant research (Oelze et al., 2016) has directly incorpo-

rated these baselines when drawing inter-site comparisons, which

instead use the offset between baseline plant isotope values and body

tissue isotope values (known as Δ13C and Δ15N) as the unit of compari-

son, thereby allowing a stronger basis upon which primate interactions

with their environment can be evaluated across sites.

Within Pan isotope ecology, particular focus has been placed on

chimpanzees in more open mosaic habitats due to their use as refer-

ential models for human evolution. To date, data from a growing

number of sites have been published (Ugalla, Ishasha: Schoeninger

et al., 1999, Gombe: Schoeninger et al., 2016, Fongoli: Sponheimer

et al., 2006, Kayan: Oelze et al., 2016, Issa: van Casteren et al.,

2018), however each employ starkly different sampling approaches,

and the majority of these studies lack baseline environmental

data. Here we describe patterns of isotopic variation in chimpanzees

inhabiting a savanna environment along a latitudinal gradient in

Senegal, at the northwestern reaches of chimpanzee distribution in

Africa. The climate of southeastern Senegal is characterized by

strong seasonality in rainfall and temperatures which consistently

exceed 40 �C, comprising exceptional conditions expected to be

particularly taxing to chimpanzees (Pruetz, 2007; Pruetz & Bertolani,

2009; Wessling, Kühl, Mundry, Deschner, & Pruetz, 2018).

This region likewise represents the northern margins of the West

African chimpanzee (Pan troglodytes verus) range. Multiple hypotheses

have been proposed to explain the potential drivers of the chimpan-

zee range margin in this region (Kortlandt, 1983; McGrew, Baldwin, &

Tutin, 1981), although food scarcity has been the predominant

(albeit untested) hypothesis as limiting chimpanzee biogeographic

distribution here (Kortlandt, 1983; Lindshield, 2014). Previous work at

Fongoli, a habituated chimpanzee research site approximately 65 km

south of the previous IUCN estimated range limit (Humle et al., 2008),

has demonstrated that chimpanzees in this region employ a range of

compensatory behaviors to cope with this environment (coprophagy:

Bertolani & Pruetz, 2011; tool-assisted hunting: Pruetz & Bertolani,

2007, Pruetz et al., 2015; pool use: Pruetz & Bertolani, 2009; cave

use: Pruetz, 2007), and appear to be physiologically challenged due to

the extremity of this environment to varying degrees (Wessling, Kühl,

et al., 2018).

Although for Fongoli chimpanzees challenges of dehydration and

thermoregulation appear more stressful than maintaining adequate

energetic status (Wessling, Kühl, et al., 2018), the relative importance

of these challenges may not remain identical for chimpanzees who

survive at the very edge of the chimpanzee range. An important ques-

tion remains as to what drives chimpanzee occurrence in an extreme

savanna habitat, and ultimately, what is most limiting to chimpanzees

inhabiting the very edges of their tolerable habitat? If environmental

conditions are assumed to be less suitable to a species the closer to

the edge of their range (Hutchinson, 1961), then we should expect

that biomarkers of the physiological challenges faced in this habitat

become more pronounced toward the range margin. Equally, if

chimpanzees at the range edge encounter increasing environmental

constraint, we should expect them also to develop adaptations for

overcoming these constraints as they are able. For example,

666 WESSLING ET AL.

individuals may increasingly rely upon the consumption of fallback

foods, or less-preferred food items eaten when preferable items

become scarce (Marshall & Wrangham, 2007), such as termites and

nonreproductive plant organs (Boesch & Boesch-Achermann, 2000;

Bogart & Pruetz, 2011; Goodall, 1986; Head, Boesch, Makaga, &

Robbins, 2011; Morgan & Sanz, 2006; Piel et al., 2017; Pruetz, 2006;

Tutin, Ham, White, & Harrison, 1997; Wrangham et al., 1991).

Here, we describe patterns of isotopic variation in chimpanzees at

the edges of the chimpanzee range limit. We expected that if main-

taining adequate nutritional and dietary requirements is increasingly

challenging with decreasing distance to the chimpanzee biogeographi-

cal range edge, then we should predict fundamental differences in iso-

topic ratios among groups at the range edge according to their

location. Alternatively, if chimpanzees are modifying their diet to

accommodate potential environmental inadequacies, dietary variation

should be visible in stable isotopic variation among these groups.

Specifically, we predict that (a) if chimpanzees are more likely to

encounter periods of starvation closer to the range margins, that

chimpanzee tissues will be δ15N enriched at the edge of the range,

and (b) that if food scarcity is a limitation to chimpanzee distribution,

then chimpanzee δ13C and δ15N isotopic profiles will indicate heavier

reliance upon less-preferred food items or fallback foods. Contingent

upon the investigation of these predictions is the hypothesis that iso-

topic baseline signatures remain largely identical across Senegal, and

that dietary differences or physiological responses to food scarcity will

be the sole origin of chimpanzee isotope variation.

2 | METHODS

We collected plant and chimpanzee hair samples along a latitudinal

gradient at the edges of the West African chimpanzee range margin in

southeastern Senegal (Figure 1). These five sites included Fongoli

(FSCP: 12�400N, 12�130W; southern-most site), a research chimpan-

zee community habituated since 2005, Kayan (12�120W 13�120N),

one of the research sites surveyed by the Pan African Programme

(Oelze et al., 2016), two unhabituated chimpanzee sites located

between Fongoli and Kayan (Kanoumering: 12�60W12�470N, Makhana:

12�70W 13�50N) as well an additional site (Hérémakhono) located

beyond the previously estimated IUCN range limit (12�00W 13�260N;

Humle et al., 2008). Surveying at and beyond the previously estimated

IUCN West African chimpanzee range limit indicated that if Héréma-

khono does not represent the very northern-most vestige where chim-

panzees can be found, it is likely very near to it (Wessling, unpublished

data). In addition, initial comparison of chimpanzee densities across

these sites suggests densities are far lower at Hérémakhono than esti-

mated at the four other sites (Wessling, unpublished data).

This region of Senegal belongs to the Sudano-Guniean vegetation

belt (Ba et al., 1997) and can be described as a savanna-woodland

mosaic comprising gallery forest, woodland, and grassland. The

climate is highly seasonal, and is particularly extreme in comparison to

habitats typical for the chimpanzee across its biogeographical range

(Wessling et al., 2018). The region undergoes a significant annual dry

season (October–April) during which little to no rain falls, and temper-

atures reach 46 �C (Wessling, Kühl, et al., 2018). Mean annual rainfall

for the region is 945 mm (FSCP, unpublished data), the majority of

which falls during the short wet season (June–August).

2.1 | Isotope analyses

We collected chimpanzee hair samples between April 2012 and

February 2014 from vacated chimpanzee night nests whenever

encountered and accessible. Samples (n = 81) were selected for analy-

sis based on sample quality (weight, length, and root present, fresh

nest, temporal resolution) of the sample, and were prepared following

the protocol outlined in Oelze et al. (2016). Hair samples ranged from

2 to 9 cm in maximum length (mean ± SD: 6.4 ± 1.5 cm), and were

aligned at the root bulb and sectioned into 2.5 mm, 5 mm, 10 mm, or

1.5 cm segments, as segment weight allowed (min. target weight:

0.2 mg). These sections were then transferred into tin capsules for

analysis at the Max Planck Institute for Evolutionary Anthropology

(MPIEVA) in Leipzig, Germany on a Flash EA 2112 (Thermo-Finnigan®,

Bremen, Germany) coupled to a DeltaXP mass spectrometer (Thermo-

Finnigan®) or on a EuroEA3000 (EuroVector SpA®, Milan, Italy)

coupled to a MAT 253 IRMS (Thermo-Finnigan®) at the German

Research Center for Environmental Health, Institute for Groundwater

Ecology, Neuherberg, Germany.

We sampled plant items from each of the aforementioned sites.

This sample set was composed of known or potential food items

for the Fongoli community (Pruetz, 2006; FSCP unpublished data).

We followed a similar plant sampling approach at the unhabituated

chimpanzee sites for which no feeding data were available. Plant sam-

ples were collected opportunistically during an initial pilot study

(April–June 2012) as well as systematically throughout the study

period (January 2013–February 2014). Sampling occurred quarterly at

three sites (Hérémakhono, Kanoumering, and Makhana), whereas

sampling was continuous at the two sites (Fongoli and Kayan) with

continuous research presence.

To control for potential canopy effects in plant samples (Cerling,

Hart, & Hart, 2004; Schoeninger et al., 1998; Van der Merwe &

Medina, 1991), we specifically targeted our sampling to the bottom of

the arboreal canopy, as low canopy samples have been suggested to

be sufficiently representative of δ13C levels in arboreal primate diet

food items (Blumenthal, Rothman, Chritz, & Cerling, 2016; Krigbaum,

Berger, Daegling, & McGraw, 2013; Nelson, 2013). Rather than

restrict our analyses to fruit items as has previously been recom-

mended (Crowley, 2012; Oelze et al., 2016), we included plant parts

from all potential Fongoli food species so as to encompass the wider

range of potential environmental isotopic contributions, but targeted

known Fongoli dietary items when available. If we were unable to

collect food items during the phenological stage at which they are

eaten at Fongoli (e.g., new leaves or ripe fruit) we collected a sample

from the species during another phenological stage (i.e., mature leaves or

unripe fruit), or in few cases, another plant organ altogether (e.g., mature

leaves instead of ripe fruit). As we were unable to directly estimate

dietary isotopic inputs, this sampling protocol provided us with an

environmental (not dietary) baseline, representative of potential (not

actual) dietary availability at each site. Kayan hair and fruit sample

isotope analyses have previously been published (Oelze et al., 2016),

although published results were averages of duplicate measurements;

WESSLING ET AL. 667

instead, here we included original single measurements in our models,

and supplemented this dataset with previously unpublished nonfruit

plant samples from Kayan. In few cases (n = 5), individual plants in our

dataset which had been sampled multiple times over the course of a

year were included in the dataset as single plant items.

Plant samples were dried using silica gel and sun exposure, and

stored dry until shipped for analysis. Prior to analysis, we dried sam-

ples in an oven at 50 �C for a minimum of 12 hr, homogenized them

using a ball mill, after which they were redried at 50 �C for a minimum

of 12 hr, and weighed into tin capsules for measurement. Plant sam-

ples were measured by three different groups: at (a) MPIEVA (see

above), in (b) Leipzig, Germany on a Flash HT Plus (Thermo Scientific®,

Waltham) coupled to a MAT 253 IRMS (Thermo-Finnigan®) in collabo-

ration with IsoDetect GmbH, or (c) on a Flash2000 Elemental Analyzer,

coupled with a Conflo IV and a Delta V Advantage Isotope Ratio Mass

Spectrometer (Thermo Fisher, Dreieich, Germany) at the University of

Leipzig. All samples from Kayan were run in Neuherberg, Germany (see

above) via partnership with IsoDetect GmbH, therefore we group these

samples within the IsoDetect GmbH analyses (see below).

We expressed the stable isotope ratios of carbon and nitrogen as13C/12C and 15N/14N ratios using the delta (δ) notation in parts per

thousand or permil (‰). This is relative to the international standard

materials Vienna PeeDee Belemite (vPDB) and atmospheric N2. We

included repetitive measurements in each run of CH7, CH6, N1, N2,

USGS 25, USGS 40, and USGS 41 and internal-laboratory standard

materials to calculate analytical error; these results suggested analyti-

cal error to be <0.3 ‰ (2σ) for both δ13C and δ15N. In general, we

aimed to analyze samples in duplicate (at minimum) and across multi-

ple laboratories whenever possible. In practice, we analyzed an aver-

age of 2.4 ± 2.4 δ15N (mean ± SD, range: 1–24) and 2.6 ± 2.5 δ13C

(range: 1–32) runs per sample, with only 54 and 14 sample results

originating from single analyses for δ15N and δ13C isotopes, respec-

tively. We observed interlaboratory analytical differences and there-

fore accounted for this discrepancy with the use of an additional term

in our models (see below).

To ensure analytical quality, we excluded all plant analyses for

which sample amplitudes were less than one-third or three times the

average amplitude of run standards for carbon (n = 112) and nitrogen

(n = 409) isotope analyses of plant samples, and four samples for hair

carbon isotope analyses. Additionally, as an extra precaution to avoid

potential biases in site averages due to sampling of extreme plant

items (e.g., C4 grasses), plant sample isotope values were excluded if

they exceeded two standard deviations from the overall mean of the

dataset (ncarbon = 80, nnitrogen = 25). We excluded all hair results

(n = 4) with atomic C:N ratios outside the 2.6–3.8 range originally

proposed by O'Connell, Hedges, Healey, and Simpson (2001). Three

additional hair δ13C results were excluded as they were extreme out-

liers (<−28‰) unlikely to be true measurements.

In total, we successfully analyzed 764 segments from 81 hair sam-

ples for δ13C and δ15N. For plant samples, we analyzed 621 runs of

265 plant samples from 53 species representing 116 unique plant

parts for δ15N, and 864 runs of 340 samples from 54 species repre-

senting 120 unique plant parts for δ13C. This included 42 food items

for which identical plant parts were collected from the same species

in at least two sites.

2.2 | Statistical analyses

To test for potential inter-site differences in baseline plant sample

isotopic values, we fitted linear mixed models (LMM) with a Gauss-

ian error function using the “lme4” package (Baayen, 2008) in R

(version 3.4.0; R Core Team, 2017) for δ15N and δ13C values

separately. Each individual plant sample analysis was included in

the model as a single data point, and controlled for ‘sample ID’ as a

random effect. We used the factor ‘site’ as our single test predictor,

and included ‘lab’ as a fixed effect to control for potential interla-

boratory differences. To account for potential seasonal variation in

plant isotopic ratios, we included a generic seasonal term (sine and

cosine of Julian date upon which each sample was collected:

Stolwijk, Straatman, & Zielhuis, 1999) as a control predictor. For

samples in which only the collection month was noted (n = 54), we

FIGURE 1 Location of the five chimpanzee research sites relative to the Pan troglodytes verus IUCN range limit (green area, Humle et al., 2008)

within southeast Senegal

668 WESSLING ET AL.

assigned these samples to the monthly midpoint (15th). Lastly, as

different plant tissues and species may differ in their isotope ratios,

we controlled for plant species and plant part as random effects,

with consideration of plant part maturity as potentially isotopically

distinct (per Blumenthal et al., 2016; Schielzeth & Forstmeier,

2009). Due to difficulty of in-field identification, we grouped two

Lannea species (L. microcarpa and L. acida), as well as several Ficus

species into a single level for each genus (with the exception of

the following species which were treated separately: sur, ingens,

cordata, sycomorus, trichopoda, umbellata, asperfolia, abutilifolia,

vallis-choudae). To keep type I error rate at the nominal 5%, we

included random slopes (Barr, Levy, Scheepers, & Tily, 2013; Schielzeth &

Forstmeier, 2009) for season within species and plant part as well as

site within species. We used a likelihood ratio test to compare the fit

of these full models to their respective null models, lacking only the

test predictor (Forstmeier & Schielzeth, 2011).

We visually inspected qq-plots and residuals plotted against fitted

values and verified that the assumptions of normally distributed and

homogeneous residuals were met in each model. These plots revealed

relatively wide (i.e., long-tailed) distributions of model residuals,

suggesting our models were possibly conservative in their ability to

identify significant effects of our test predictor. Further, there was a

slight positive correlation between model residuals and fitted values,

which is likely due to a large number of random effect levels occur-

ring relatively rarely in the dataset and leading to more prominent

“shrinkage” (Baayen, 2008). We evaluated model stability by exclud-

ing levels of the random effects one at a time and comparing the

estimates derived from these datasets with those derived for the full

dataset. This revealed no influential random effects. We verified our

models did not suffer from collinearity among predictors by deter-

mining Variance Inflation Factors (VIF; Field, 2005) for a standard

linear model excluding all random effects (maximum VIF: 1.07).

We chose not to test for statistical differences between sites for

chimpanzee hair sample analyses due to our inability to account for

potential pseudo-replication arising from multiple sampling of the

same individuals in the unhabituated sites (Mundry & Oelze, 2016).

Instead, we describe broad-scale comparison among these groups

using only site averages and δ13C and δ15N ranges. We did, however,

calculate site averages using an LMM for each site and isotope which

controlled for the random effect of “sample ID”. Lastly, we calculated

isotopic fractionation (Δ) following Oelze et al. (2016) to control for

plant baseline effects and to allow for inter-site dietary comparison. In

brief, isotope fractionation values, or Δ13C and Δ15N, are calculated as

the difference between average site hair isotope values and average

baseline plant values (using intercept and estimates per site in our plant

models). All research was noninvasive and was approved by the Max

Planck Society.

3 | RESULTS

Our model reported no significant differences in plant δ13C ratios

among the five sites (full-null model comparison: χ2 = 3.986, df = 4,

p = .408), indicating Hérémakhono (as the reference category) was

not significantly higher in plant δ13C values than the four southern

sites (Fongoli, Kanoumering, Makhana, Kayan; Table 1(a)) when con-

trolling for potential biases among the sites. δ15N significantly differed

among the five sites (χ2 = 23.637, df = 4, p < .001), with plant δ15N

values of Hérémakhono baseline plant samples significantly higher

than in the four other sites (Table 1(b)). Model-based site δ13C plant

averages ranged between −28.7 and −29.2‰ across the five sites,

whereas the four southern-most sites averaged between 0.9‰ and

1.6‰ lower than Hérémakhono (3.5‰) in δ15N plant values (Table 2).

We also found clear differences between the southern sites and

Hérémakhono in mean chimpanzee hair isotope values (Table 3;

Figure 2). Average δ13C and δ15N values for the southern communi-

ties ranged from −23.1 to −22.6‰ and 3.2 to 3.6‰, respectively,

whereas Hérémakhono samples averaged at least 0.9‰ higher in

δ13C and 1.0‰ higher in δ15N than even their most similar counter-

parts. The southern sites exhibited very minor overlap with the range

of isotopes observed at Hérémakhono. In general, intra-site variation

in hair δ13C and δ15N values was greater than inter-site average vari-

ation, with intra-site variation ranging from 1.2‰ to 2.6‰ in δ13C

and 1.1‰ to 2.4‰ in δ15N among the five sites, with Fongoli exhi-

biting the highest variation in both isotopes while likewise being the

most heavily sampled site by at least twofold.

By accounting for inter-site variation in baseline plant samples at

each site, Δ13C and Δ15N values were tightly clustered for all five sites

relative to total Pan variation (Figure 3), averaging 6.3‰ for Δ13C

(range: 5.8‰ to 7.2‰) and 1.2‰ in Δ15N (range: 1.0‰ to 1.6‰)

across sites. Maximum difference between Δ values across sites were

TABLE 1 Model results for the effect of ‘site’ on δ13C (a) and δ15N(b) ratios in plant samples

(a) δ13CTerm Estimate ± SE p-value

(Intercept) −28.88 ± 0.42 –

Site (Fongoli)a 0.19 ± 0.30

Site (Kanoumering)a −0.24 ± 0.39 0.377

Site (Kayan)a −0.28 ± 0.28

Site (Makhana)a 0.13 ± 0.35

Sine (Julian date) 0.36 ± 0.20 –

Cosine (Julian date) −0.26 ± 0.23 –

Lab (MPI)b 0.10 ± 0.07 –

Lab (UNI)b −0.69 ± 0.10 –

(b) δ15NTerm Estimate ± SE p-value

(Intercept) 3.55 ± 0.37 –

Site (Fongoli)a −1.37 ± 0.33

Site (Kanoumering)a −1.62 ± 0.36 <0.001

Site (Kayan)a −1.62 ± 0.39

Site (Makhana)a −1.10 ± 0.40

Sine (Julian date) −0.12 ± 0.12 –

Cosine (Julian date) 0.13 ± 0.17 –

Lab (MPI)b −0.13 ± 0.14 –

Lab (UNI)b −2.11 ± 0.22 –

a Estimate refers to the comparison with Hérémakhono as the referencecategory.

b Estimate refers to the comparison with IsoDetect as the reference category.

WESSLING ET AL. 669

1.4‰ and 0.6‰ for Δ13C and Δ15N respectively, with highest Δ

values at Hérémakhono (Δ13C) and Kanoumering (Δ15N).

Lastly, Figure 4 demonstrates the variability of different plant

food items within our dataset, across all species and sites, and rela-

tive to the modeled averages of each plant part controlling for sam-

pling biases. Broadly, δ15N demonstrated little variation among plant

parts with the exception of bark which was 15N depleted (although

the model failed to account for this variability). Conversely, δ13C was

relatively more variable across plant parts, with mature leaves, new

leaves, and bark averaging lower than ripe and unripe fruit. Flowers

and pith averaged higher than fruit in δ13C, but these differences

were not reflected in the random intercepts for these parts. None-

theless, all plant parts with the exception of bark (sampled from only

two species) varied across the range of values observed for δ13C and

δ13N plant analyses, indicating strong interspecific variability in these

isotope systems.

4 | DISCUSSION

Altogether, isotope ratios for Fongoli samples were similar to previ-

ously published results on the same population (Loudon et al.,

2016; Sponheimer et al., 2006), both in hair and plant datasets,

despite differences between datasets in analyses, sample size, and

dietary coverage. Our larger plant sample set indicated a wider

range of potential isotopic δ13C and δ15N inputs than previously

published estimates lacking information on the nature of the

plant samples (Loudon et al., 2016), and approach ranges similar to

that seen in wetter habitats (Carlson & Crowley, 2016; Carlson &

Kingston, 2014; Oelze et al., 2016). Nonetheless, inter-study simi-

larity in Fongoli hair isotope averages indicated interannual dietary

fidelity at Fongoli.

Generally, we found that chimpanzees at all five sites exhibited

relatively high Δ13C and low Δ15N values in comparison to chimpan-

zees from other habitat types. Our results support that low δ15N

values in this region are due in part to equally low environmental

baseline levels (Sandberg et al., 2012), and contradict assertions that

dry African sites reveal δ15N enriched values (Heaton, Vogel, von La

Chevallerie, & Collett, 1986). In addition, Δ15N values for all sites

averaged lower than would be expected for a trophic level increase

between a consumer and its diet (~3‰; Minagawa & Wada, 1984).

Low Δ15N values in Senegal have previously been suggested either to

be due to relatively low meat consumption, or to result from the regu-

lar consumption of 15N depleted Macrotermes termites (n = 4, mean

0.2‰ δ15N, Oelze, unpublished data; Oelze et al., 2016, Sandberg

et al., 2012). Mound-building termites are abundant (Bogart & Pruetz,

2011) and ubiquitous in the study region (Wessling, personal observa-

tion), and are consumed at higher rates at Fongoli than other chimpan-

zee sites (Bogart & Pruetz, 2011). Termite remains in Kayan fecal

samples (Oelze et al., 2016) indicate termite consumption across multi-

ple sites in Senegal, thereby supporting 15N-depleted termite consump-

tion as a potential cause for low Δ15N across our sites. Additionally,

Fongoli chimpanzees appear to consume meat less frequently, with

only 99 hunts observed over 9 years (Pruetz et al., 2015), relative to

other chimpanzee communities such as Taï where hunts are observed

on average every 4 days (Samuni et al., 2018), or Ngogo with an aver-

age 73 hunts per year (Watts & Mitani, 2002). If low hunting rates is a

region-wide pattern, comparatively infrequent meat consumption may

also explain lower Δ15N values in our samples relative to sites like Ngogo

and Taï. Nonetheless, direct behavioral comparison is needed to confirm

potential drivers of low Δ15N in Senegalese chimpanzee hair samples.

As for chimpanzee isotopic variation across our five sites, we

found strong similarities in δ13C and δ15N hair sample averages and

significant overlap in the ranges of these isotope ratios among the

TABLE 2 Plant baseline δ13C and δ15N values from five sites in Senegal, ordered in table from south (top) to north (bottom), and their descriptive

statistics for 340 (δ13C) and 265 (δ15N) samples

Site n13C n15Nδ13Ca

(model-based; x ± SE) δ13Cb (x ± SE)δ13C minimum,maximum

δ15Na

(model-based; x ± SE) δ15Nb (x ± SE)δ15N minimum,maximum

Fongoli 154 122 −28.7 −28.6 ± 0.4 −32.4, −24.0 2.2 2.0 ± 0.2 −3.0, 7.2

Kanoumering 38 35 −29.1 −28.4 ± 0.4 −32.1, −24.6 1.9 1.9 ± 0.3 −1.6, 7.3

Makhana 46 33 −28.7 −28.3 ± 0.4 −32.3, −24.6 2.4 2.1 ± 0.3 −1.8, 6.4

Kayan 69 51 −29.2 −28.8 ± 0.3 −32.3, −24.6 1.9 1.9 ± 0.7 −1.7, 5.3

Hérémakhono 35 26 −28.9 −28.8 ± 0.3 −32.3, −24.9 3.5 3.2 ± 0.4 −0.9, 7.2

a Model-based averages derived from intercept and estimates of respective LMMs containing the full dataset.b Controlling for species, plant part, sample, season, and lab and the random slopes of season within plant part and species, within each site.

TABLE 3 Descriptive statistics for 81 chimpanzee hair sample δ13C and δ15N values from five sites located along a latitudinal gradient in Senegal,

ordered in table from south (top) to north (bottom), as well as site mean Δ13C and Δ15N in comparison with plant baselines per site

Site nsamples nsegments δ13Ca (mean ± SE) δ13C minimum, maximum δ15Na (mean ± SE) δ15N minimum, maximum Δ13Cb Δ15Nb

Fongoli 37 378 −22. 6 ± 0.1 −23.5, −20.9 3.2 ± 0.1 1.9, 4.3 6.1 1.0

Kanoumering 16 145 −22.8 ± 0.1 −23.7, −22.3 3.5 ± 0.1 2.7, 4.8 6.3 1.6

Makhana 11 110 −22.9 ± 0.1 −23.5, −22.1 3.6 ± 0.0 3.0, 4.1 5.8 1.2

Kayan 10 89 −23.1 ± 0.1 −23.9, −22.3 3.2 ± 0.1 2.4, 4.2 6.1 1.3

Hérémakhono 7 43 −21.7 ± 0.1 −22.2, −21.0 4.6 ± 0.1 3.8, 5.3 7.2 1.1

a Controlling for hair sample.b Delta values calculated using model-based plant estimates and site hair averages.

670 WESSLING ET AL.

four southern communities. However, chimpanzees from our north-

ernmost site, Hérémakhono, demonstrated δ13C and δ15N values

around 1‰ higher than their southern counterparts. In absence of

corresponding environmental baseline data, these results might have

been suggestive of significant differences among the communities

either in dietary composition, energetic status, or both.

However, calculations of Δ13C and Δ15N allow us to control for plant

baselines and turn our attention to the dietary behavior or other potential

physiologically-based influence on isotopic variation that we sought to

assess in this study. By accounting for the observation that Hérémakhono

plant samples averaged 1.4‰ in δ15N higher than the southern groups,

we identified that Δ15N values were functionally identical across sites

with a maximum 0.6‰ variation for Δ15N among the site averages.

Therefore, regardless of potential sources of variation in δ15N values

of environmental baselines, the scale of differences in plant baselines

among these sites accounted for the majority of isotopic differences

in δ15N values in our hair samples. The low degree of Δ15N variation

between sites is contradictory to our predictions that chimpanzees

at Hérémakhono experience more pronounced periods of nutritional

stress than their counterparts. However, as plant δ13C values were

statistically indistinguishable across sites, the mean Hérémakhono

Δ13C value (7.0‰) remained higher than its most similar counterpart

by 0.9‰, twice the maximum difference in averages among the four

southern sites (0.4‰). Taken together, as δ13C differences between

chimpanzees from different sites are not explained by differences in

environmental baseline, dietary rather than direct environmental iso-

topic differences must account for differences in Δ13C, with the diet

of Hérémakhono chimpanzees proving to be relatively 13C enriched.

Without direct behavioral evidence, identification of dietary

explanations for 13C enrichment at Hérémakhono remains speculative,

although a few possible candidates emerge from our data. Within our

sample set, the most likely candidate explaining 13C enrichment could

be a greater reliance on flowers at Hérémakhono than at other sites, as

flowers were the only plant food type which averaged higher in δ13C

across species than ripe fruits (although all random intercepts in this

model fell below the ripe fruit average). It is also possible that specific13C enriched food items within this dataset (e.g., Afzelia or Parkia fruits)

could also be candidates for overall dietary 13C enrichment, but that

this was not a consistent pattern among plant parts and species.

Another common explanation for 13C enrichment is the consump-

tion of C4 plants. Fongoli chimpanzees have been observed to infre-

quently consume C4 plants like grasses (Pruetz, 2006), therefore it could

also be possible that greater consumption of C4 grasses may explain13C enrichment at Hérémakhono. Furthermore, typical C4 crop spe-

cies like corn and sugarcane are grown within the study region,

although there have been no reports of crop raiding of these species

from chimpanzees at any of our sites. Yet, chimpanzees at Héréma-

khono have been exclusively reported to annually crop-raid mango

trees located within corn fields in the months of May and June

(Wessling, pers. observation). Although we do not have isotopic data

FIGURE 3 Average Δ13C and Δ15N values per Pan research site; size of

points denotes relative sizes of hair datasets (range: 6–36). Δ values arecalculated based on the following hair and plant datasets: Oelze et al.,2016; Loudon et al., 2016; van Casteren et al., 2018; this study

FIGURE 2 (a) δ13C and δ15N hair sample averages for five locations along a north–south gradient in Senegal, with ranges of all hair segments for

each site. Sample means (points) and site ranges (polygons) are depicted for Fongoli (red), Makhana (yellow), Kanoumering (blue), Kayan (green),and Hérémakhono (purple). (b) Hair sample averages controlled for site plant sample baselines (Δ13C and Δ15N)

WESSLING ET AL. 671

from mango fruits, it is likely that these fruits are 13C enriched if the

fields containing them are irrigated (Wallace et al., 2013), or are sunnier

than locations where wild plant foods were sampled (e.g., France, 1996).

Regardless, isotope systems are highly complex and sources of

variation are multiplicitous. Therefore, conclusions in wild great ape

isotope ecology should remain relatively tentative until supported

with behavioral consensus and more detailed physiological studies.

Although consistent patterns among plant parts and species have

been described, the complexities of identifying origins of isotopic vari-

ation in Pan diets should promote researchers to be cautious in their

interpretations and coarse in their conclusions.

Previous comparison across habitats has established that food avail-

ability in savanna environments is scarcer than more forested habitats

(Wessling, Deschner, et al., 2018). It may therefore be possible that food

scarcity is a real threat to chimpanzees at the edge of their range, but that

they are able to employ flexible strategies to counteract this threat or

dampen its effects below a measureable threshold (Kempster et al.,

2007). Data from the only site for which feeding strategy is known,

the habituated community of Fongoli, suggests that although savanna

chimpanzees are ripe-fruit specialists, they depend to a greater extent on

fallback foods when preferred ripe fruit items are not available (Bogart &

Pruetz, 2011; Pruetz, 2006). As the four unhabituated chimpanzee com-

munities demonstrated similar Δ15N ratios to Fongoli, it is likely this is a

ubiquitous but necessary strategy employed by these chimpanzees to

cope with a seasonal and challenging environment.

However, we did not observe a measureable difference in Δ15N

values for Hérémakhono chimpanzees, which suggests that these

chimpanzees also do not endure periods of starvation or energetic

stress to greater degrees than the southern communities. Such a strat-

egy would only lead to low inter-site variation in diet where dietary

pressures are equal in their magnitude. However, if floristic diversity

and/or density are reduced at Hérémakhono compared to the south-

ern sites, it may be possible that Hérémakhono chimpanzees are more

reliant upon fallback foods like flowers, grasses, or cultivated plants to

compensate for potential dietary shortcomings relative to their south-

ern counterparts. Such a strategy may explain why we observed clear

differences in Hérémakhono chimpanzee Δ13C values, but not in

Δ15N values. Yet, although laboratory experiments on bonobos

(Deschner et al., 2012) and other taxa (McCue & Pollock, 2008) have

demonstrated 15N enrichment due to nutritional stress, it may be pos-

sible that energetic deficits in situ are not easily identifiable using iso-

topic data. In this vein, Vogel et al. (2011) failed to find a δ15N effect

of nutritional scarcity in wild orangutans, even when other markers of

energetic deficit were observed. Therefore, we offer this conclusion

with caution, and acknowledge the need of further investigation into

potential ecological differences and their potential physiological con-

sequences among these sites.

Nonetheless, inter-site differences in average δ13C and δ15N

values were smaller than intra-site variation among hair segments, and

total observed δ13C and δ15N intra-site variation across all five com-

munities are comparable to the range of variation in other chimpanzee

communities across Africa (Oelze et al., 2016). It is notable, however,

that δ13C and δ15N hair ranges for these sites varied considerably

and extended across the range of observed variation for Pan sites

(e.g., Oelze et al., 2016), suggesting the isotopic composition of chim-

panzee diets in these areas varies substantially within the year—

further conforming to behavioral observation that the Fongoli diet

flexibly responds to availability of various food items, and relies more

heavily upon fallback food items when necessary. Such relatively tight

grouping in Δ13C and Δ15N site averages may therefore suggest that

chimpanzees at all sites have developed similar dietary adaptations for

coping with reduced food availability when encountered, and that this

strategy extends across the range of food scarcity encountered in our

study area, albeit to varying degrees.

It is worth noting that the variability in site ranges suggests that

the extent of intra-annual dietary variation could also vary from site

to site, although such inference should remain cautious as isotopic

ranges per site appeared to increase with increasing sample size. Fur-

ther investigations into intra-annual variation in isotopic values and

small-scale ecological variation will illuminate the flexibility afforded

to individuals living in relatively food-scarce environments (Wessling,

Deschner, et al., 2018).

Lastly, this research underlines the necessity of accounting for

potential baseline differences in isotope ecology of a species when

engaging in inter-site comparisons. Multiple authors (e.g., Carlson &

Crowley, 2016; Oelze et al., 2016) have previously proposed this

approach, although conclusions continue to be drawn from wild ape

isotope analyses without incorporation of environmental baselines

(e.g., Loudon et al., 2016; Schoeninger et al., 2016). Our research

highlights that such baselines are necessary not only across varied

FIGURE 4 Plant part variation in δ13C (top) and δ15N (bottom) values.

Shown are medians (thick horizontal lines; “MED”), quartiles (boxes),percentiles (vertical lines; 2.5 and 97.5%), and random effectintercepts (red diamond; “REI”) of each plant part (left to right: ripefruit, unripe fruit, mature leaves, new leaves, bark, flower, and pith)

672 WESSLING ET AL.

habitats, but also within habitats which might appear superficially

similar. In absence of environmental baselines, our understanding

and interpretation of isotopic variation in our data would have led to

inaccurate conclusions. For example, the lack of plant baseline data

in our research would have potentially led us to conclude that

enrichment of δ15N levels in Hérémakhono chimpanzees potentially

stems from greater periods of starvation due to food scarcity at the

extreme edge of chimpanzee distribution relative to their southern

counterparts. Given that Hérémakhono chimpanzees occur at low

densities (Wessling, unpublished data) and are likely inhabiting the

extreme edges of tolerable chimpanzee habitat, such a conclusion

from our data may not have been implausible.

We argue further that accounting for environmental baselines in

inter-site comparison will likewise remain ineffective if potential

sources of isotopic variation within unbalanced plant baseline datasets

are ignored. Typically, baseline values are calculated on a site-by-site

basis as the average of all plant samples analyzed at each site, and

usually sample sets are intentionally limited to certain components of

the diet (e.g., ripe fruit: Loudon et al., 2016; Oelze et al., 2016). How-

ever, by expanding the dataset to encompass the wider range of the

chimpanzee diet, while accounting for the effects of plant species,

plant part, and season across the entire comparative dataset, we were

able to distinguish patterns of differentiation in environmental base-

lines across our dataset which likewise would have been overlooked

had we not benefited from the explanatory power of mixed models.

Again, plant baseline averages appeared to be nearly identical in δ13C

values across sites when averages were created site-by-site: instead

controlling for plant species and plant part in the entire dataset helped

distinguish inter-site differences otherwise masked by biases in sam-

pling across sites. In consideration of these methodological effects,

these results therefore underline that future primate isotopic research

should be as comprehensive in scope as feasibly possible and unwa-

veringly careful in interpretation. The implications of our methods

likewise extend beyond nonhuman primates to humans and other

taxa, and highlight that caution is needed in interpretation of all

comparative isotopic datasets when environmental baselines cannot

be established.

5 | CONCLUSION

Our results show that Hérémakhono, likely the northern-most chim-

panzee community on the continent and at the physical edge of the

West African chimpanzee range, had distinctly higher δ13C and δ15N

values relative to their counterparts further south in Senegal. How-

ever, baseline plant δ15N values also were significantly higher than

those from the four southern sites, thereby suggesting nitrogen frac-

tionation was similar for chimpanzees across all five Senegalese sites.

These results support two main conclusions. First, that Hérémakhono

chimpanzees are likely addressing potential dietary constraints using13C enriched food items at the range edge as they have a measur-

ably δ13C enriched diet compared to the four southern sites. In doing

so, it appears that they are able to avoid potentially higher risks of

starvation relative to their southern counterparts to below a mea-

sureable threshold using dietary flexibility. Our second conclusion is

methodological; due to the numerous and varied ways in which local

environments can impact an individual's isotopic profile, accounting

for baseline environmental profile while concurrently accounting for

sources of bias within each baseline is compulsory when comparing

geographically distinct groups. This remains pertinent even when

absolute distances between comparative groups may be small (in this

case <100 km), or when habitats would appear to be superficially

similar. By coupling environmental baseline data with our interpreta-

tion of hair isotopic variation, we demonstrate dietary similarities

across five chimpanzee communities living at the limits of their bio-

geographical range.

ACKNOWLEDGMENTS

We thank the Republic of Senegal and the Département des Eaux et

Forêts, et de Chasses for their permission to conduct this work. For

their assistance in coordination of the Pan African Programme we

thank Mimi Arandjelovic, Paula Dieguez, and Claudia Herf. For assis-

tance in the field we thank Jacques Tamba Keita, Kaly Bindia, Madi

Keita, Dondo Kante, and Michel Sadiakho. For their assistance with

analyses we thank Michael Richards, Annabelle Reiner, Ulrike

Wacker, Sven Steinbrenner, and the staff at IsoDetect GmbH, as well

as Klervia Jaouen for her invaluable support in the process. We

thank Colleen Stephens for her assistance with figure development.

Sincere thanks to Liran Samuni, Christophe Boesch, Roger Mundry,

and two anonymous reviewers for their helpful comments on earlier

versions of this manuscript. This research was funded by the Max

Planck Society, the National Geographic Society, Iowa State Univer-

sity, the Max Planck Society Innovation Fund, and the Krekeler

Foundation.

ORCID

Erin G. Wessling https://orcid.org/0000-0001-9661-4354

REFERENCES

Ba, A., Sambou, B., Ervik, F., Goudiaby, A., Camara, C., & Diallo, D. (1997).Végétation et flore. Parc Transfrontalier du Niokolo Badiar. Union Eur-opéenne-Niokolo Badiar, ISE.

Baayen, R. H. (2008). Analyzing linguistic data: A practical introduction tostatistics using R. Cambridge: Cambridge University Press.

Balasse, M., Bocherens, H., Mariotti, A., & Ambrose, S. H. (2001). Detec-tion of dietary changes by intra-tooth carbon and nitrogen isotopicanalysis: An experimental study of dentine collagen of cattle (Bos taurus).Journal of Archaeological Science, 28(3), 235–245.

Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effectsstructure for confirmatory hypothesis testing: Keep it maximal. Journalof Memory and Language, 68(3), 255–278.

Bertolani, P., & Pruetz, J. D. (2011). Seed reingestion in savannah chimpan-zees (Pan troglodytes verus) at Fongoli, Senegal. International Journal ofPrimatology, 32(5), 1123–1132.

Blumenthal, S. A., Rothman, J. M., Chritz, K. L., & Cerling, T. E. (2016). Sta-ble isotopic variation in tropical forest plants for applications in prima-tology. American Journal of Primatology, 78(10), 1041–1054.

Boesch, C., & Boesch-Achermann, H. (2000). The chimpanzees of the TaïForest: Behavioural ecology and evolution. New York: Oxford UniversityPress, USA.

Bogart, S. L., & Pruetz, J. D. (2011). Insectivory of savanna chimpanzees(Pan troglodytes verus) at Fongoli, Senegal. American Journal of PhysicalAnthropology, 145(1), 11–20.

WESSLING ET AL. 673

Caraveo-Patino, J., Hobson, K. A., & Soto, L. A. (2007). Feeding ecology ofgray whales inferred from stable-carbon and nitrogen isotopic analysisof baleen plates. Hydrobiologia, 586(1), 17–25.

Carlson, B. A., & Crowley, B. E. (2016). Variation in carbon isotope valuesamong chimpanzee foods at Ngogo, Kibale National Park and Bwindiimpenetrable National Park, Uganda. American Journal of Primatology,78(10), 1031–1040.

Carlson, B. A., & Kingston, J. D. (2014). Chimpanzee isotopic ecology: Aclosed canopy C3 template for hominin dietary reconstruction. Journalof Human Evolution, 76, 107–115.

Carter, M. L. (2001). Sensitivity of stable isotopes (13C, 15N, and 18O) in boneto dietary specialization and niche separation among sympatric primatesin Kibale National Park, Uganda. University of Chicago, Department ofAnthropology, Chicago, IL, USA.

Cerling, T. E., Hart, J. A., & Hart, T. B. (2004). Stable isotope ecology in theIturi Forest. Oecologia, 138(1), 5–12.

Codron, D., Lee-Thorp, J. A., Sponheimer, M., de Ruiter, D., & Codron, J.(2006). Inter-and intrahabitat dietary variability of chacma baboons(Papio ursinus) in South African savannas based on fecal δ13C, δ15N,and % N. American Journal of Physical Anthropology, 129(2), 204–214.

Codron, D., Lee-Thorp, J. A., Sponheimer, M., De Ruiter, D., & Codron, J.(2008). What insights can baboon feeding ecology provide for earlyhominin niche differentiation? International Journal of Primatology, 29(3), 757–772.

Crowley, B. E. (2012). Stable isotope techniques and applications for pri-matologists. International Journal of Primatology, 33(3), 673–701.

Crowley, B. E., Rasoazanabary, E., & Godfrey, L. R. (2014). Stable isotopescomplement focal individual observations and confirm dietary variabilityin reddish-gray mouse lemurs (Microcebus griseorufus) from southwesternMadagascar. American Journal of Physical Anthropology, 155(1), 77–90.

DeNiro, M. J., & Epstein, S. (1978). Influence of diet on the distributionof carbon isotopes in animals. Geochimica et Cosmochimica Acta, 42(5), 495–506.

Deschner, T., Fuller, B. T., Oelze, V. M., Boesch, C., Hublin, J. J.,Mundry, R., … Hohmann, G. (2012). Identification of energy consump-tion and nutritional stress by isotopic and elemental analysis of urine inbonobos (Pan paniscus). Rapid Communications in Mass Spectrometry,26(1), 69–77.

Fahy, G. E., Boesch, C., Hublin, J. J., & Richards, M. P. (2015). The effec-tiveness of using carbonate isotope measurements of body tissues toinfer diet in human evolution: Evidence from wild western chimpan-zees (Pan troglodytes verus). Journal of Human Evolution, 88, 70–78.

Fahy, G. E., Richards, M., Riedel, J., Hublin, J.-J., & Boesch, C. (2013). Stableisotope evidence of meat eating and hunting specialization in adultmale chimpanzees. Proceedings of the National Academy of Sciences,110(15), 5829–5833.

Field, A. (2005). Discovering statistics using SPSS (Vol. 2), Thousand Oaks,California: Sage Publications.

Forstmeier, W., & Schielzeth, H. (2011). Cryptic multiple hypotheses test-ing in linear models: Overestimated effect sizes and the winner's curse.Behavioral Ecology and Sociobiology, 65(1), 47–55.

France, R. (1996). Carbon isotope ratios in logged and unlogged boreal for-ests: Examination of the potential for determining wildlife habitat use.Environmental Management, 20(2), 249–255.

Fuller, B. T., Fuller, J. L., Sage, N. E., Harris, D. A., O'Connell, T. C., &Hedges, R. E. (2005). Nitrogen balance and δ15N: Why you're not whatyou eat during nutritional stress. Rapid Communications in Mass Spec-trometry, 19(18), 2497–2506.

Goodall, J. (1986). The chimpanzees of Gombe: Patterns of behavior. Cam-bridge, MA: The Belknap Press of Harvard University Press.

Harrison, S. M., Zazzo, A., Bahar, B., Monahan, F. J., Moloney, A. P.,Scrimgeour, C. M., & Schmidt, O. (2007). Using hooves for high-resolution isotopic reconstruction of bovine dietary history. RapidCommunications in Mass Spectrometry, 21(4), 479–486.

Head, J., Boesch, C., Makaga, L., & Robbins, M. (2011). Sympatric chimpan-zees and gorillas in Loango National Park, Gabon: Dietary composition,seasonal changes and inter-site comparisons. International Journal ofPrimatology, 32, 755–775.

Heaton, T. H., Vogel, J. C., von La Chevallerie, G., & Collett, G. (1986). Cli-matic influence on the isotopic composition of bone nitrogen. Nature,322(6082), 822–823.

Humle, T., Boesch, C., Duvall, C., Ellis, C. M., Farmer, K. H., Herbinger, I., …Oates, J. F. (2008). Pan troglodytes ssp. verus. The IUCN Red List ofThreatened Species 2008.

Hutchinson, G. E. (1961). The paradox of the plankton. American Naturalist,95(882), 137–145.

Kempster, B., Zanette, L., Longstaffe, F. J., MacDougall-Shackleton, S. A.,Wingfield, J. C., & Clinchy, M. (2007). Do stable isotopes reflect nutri-tional stress? Results from a laboratory experiment on song sparrows.Oecologia, 151(3), 365–371.

Koch, P. L., Heisinger, J., Moss, C., Carlson, R. W., Fogel, M. L., &Behrensmeyer, A. K. (1995). Isotopic tracking of change in diet andhabitat use in African elephants. Science, 267(5202), 1340–1343.

Kohn, M. J. (1999). You are what you eat. Science, 283(5400), 335–336.Kortlandt, A. (1983). Marginal habitats of chimpanzees. Journal of Human

Evolution, 12(3), 231–278.Krigbaum, J., Berger, M. H., Daegling, D. J., & McGraw, W. S. (2013). Stable

isotope canopy effects for sympatric monkeys at Taï Forest, CôteD'ivoire. Biology Letters, 9(4), 20130466.

Kühl, H. S., Sop, T., Williamson, E. A., Mundry, R., Brugière, D., Campbell, G.,… Herbinger, I. (2017). The critically endangered western chimpanzeedeclines by 80%. American Journal of Primatology, 79(9), e22681.

Lee, S. H., Schell, D. M., McDonald, T. L., & Richardson, W. J. (2005).Regional and seasonal feeding by bowhead whales Balaena mysticetusas indicated by stable isotope ratios. Marine Ecology Progress Series,285, 271–287.

Lindshield, S. M. (2014). Multilevel analysis of the foraging decisions of west-ern chimpanzees (Pan troglodytes verus) and resource scarcity in asavanna environment at Fongoli, Senegal (Doctoral Thesis). Iowa StateUniversity, Ames, IA.

Loudon, J. E., Sandberg, P. A., Wrangham, R. W., Fahey, B., & Sponheimer, M.(2016). The stable isotope ecology of Pan in Uganda and beyond. Ameri-can Journal of Primatology, 78(10), 1070–1085.

Macho, G. A. (2016). From rainforests to savannas and back: The impact ofabiotic factors on non-human primate and hominin life histories. Qua-ternary International, 448, 5–13.

Macho, G. A., & Lee-Thorp, J. A. (2014). Niche partitioning in sympatricGorilla and Pan from Cameroon: Implications for life history strategiesand for reconstructing the evolution of hominin life history. PLoS One,9(7), e102794.

Marshall, A. J., & Wrangham, R. W. (2007). Evolutionary consequences offallback foods. International Journal of Primatology, 28(6), 1219–1235.

McCue, M. D., & Pollock, E. D. (2008). Stable isotopes may provide evi-dence for starvation in reptiles. Rapid Communications in Mass Spec-trometry, 22(15), 2307–2314.

McGrew, W. C., Baldwin, P. J., & Tutin, C. E. (1981). Chimpanzees in a hot,dry and open habitat: Mt. Assirik, Senegal, West Africa. Journal ofHuman Evolution, 10(3), 227–244.

Minagawa, M., & Wada, E. (1984). Stepwise enrichment of 15N along foodchains: Further evidence and the relation between δ15N and animalage. Geochimica et Cosmochimica Acta, 48(5), 1135–1140.

Morgan, D., & Sanz, C. (2006). Chimpanzee feeding ecology and compari-sons with sympatric gorillas in the Goualougo Triangle, Republic ofCongo. Cambridge Studies in Biological Studies and Evolutionary Anthro-pology, 48, 97.

Mundry, R., & Oelze, V. M. (2016). Who is who matters—The effects ofpseudoreplication in stable isotope analyses. American Journal of Prima-tology, 78(10), 1017–1030.

Nelson, S. V. (2013). Chimpanzee fauna isotopes provide new interpreta-tions of fossil ape and hominin ecologies. Proceedings of the Royal Soci-ety of London B: Biological Sciences, 280(1773), 20132324.

O'Connell, T. C., Hedges, R. E. M., Healey, M. A., & Simpson, A. H. R. W.(2001). Isotopic comparison of hair, nail and bone: Modern analyses.Journal of Archaeological Science, 28(11), 1247–1255.

Oelze, V. M., Fahy, G., Hohmann, G., Robbins, M. M., Leinert, V., Lee, K., …Kuhl, H. S. (2016). Comparative isotope ecology of African great apes.Journal of Human Evolution, 101, 1–16.

Oelze, V. M., Fuller, B. T., Richards, M. P., Fruth, B., Surbeck, M.,Hublin, J. J., & Hohmann, G. (2011). Exploring the contribution and sig-nificance of animal protein in the diet of bonobos by stable isotoperatio analysis of hair. Proceedings of the National Academy of Sciences,108(24), 9792–9797.

674 WESSLING ET AL.

Oelze, V. M., Head, J. S., Robbins, M. M., Richards, M., & Boesch, C.(2014). Niche differentiation and dietary seasonality among sympatricgorillas and chimpanzees in Loango National Park (Gabon) revealed bystable isotope analysis. Journal of Human Evolution, 66, 95–106.

Piel, A. K., Strampelli, P., Greathead, E., Hernandez-Aguilar, A., Moore, J. J., &Stewart, F. A. (2017). The diet of open-habitat chimpanzees (Pan troglo-dytes schweinfurthii) in the Issa valley, western Tanzania. Journal of HumanEvolution, 112, 57–69.

Pruetz, J. D. (2006). Feeding ecology of savanna chimpanzees (Pan troglo-dytes verus) at Fongoli, Senegal. In G. Hohmann, M. Robbins, & C. Boesch(Eds.), Feeding ecology in apes and other primates (pp. 161–182).Cambridge: Cambridge University Press.

Pruetz, J. D. (2007). Evidence of cave use by savanna chimpanzees (Pantroglodytes verus) at Fongoli, Senegal: Implications for thermoregulatorybehavior. Primates, 48(4), 316–319.

Pruetz, J. D., & Bertolani, P. (2007). Savanna chimpanzees, Pan troglodytesverus, hunt with tools. Current Biology, 17, 1–6.

Pruetz, J. D., & Bertolani, P. (2009). Chimpanzee (Pan troglodytes verus)behavioral responses to stresses associated with living in a savannah-mosaic environment: Implications for hominin adaptations to openhabitats. Paleoanthropology, 2009, 252–262.

Pruetz, J. D., Bertolani, P., Boyer Ontl, K., Lindshield, S., Shelley, M., &Wessling, E. G. (2015). New evidence on the tool-assisted huntingexhibited by chimpanzees (Pan troglodytes verus) in a savannah habitatat Fongoli, Sénégal. Royal Society Open Science, 2(4), 140507.

R Core Team. (2017). R: A language and environment for statistical comput-ing. Vienna, Austria: R Foundation for Statistical Computing.

Samuni, L., Preis, A., Deschner, T., Crockford, C., & Wittig, R. M. (2018).Reward of labor coordination and hunting success in wild chimpanzees.Communications Biology, 1, 1–9.

Sandberg, P. A., Loudon, J. E., & Sponheimer, M. (2012). Stable isotopeanalysis in primatology: A critical review. American Journal of Primatol-ogy, 74(11), 969–989.

Schielzeth, H., & Forstmeier, W. (2009). Conclusions beyond support: Over-confident estimates in mixed models. Behavioral Ecology, 20(2), 416–420.

Schoeninger, M. J. (2009). δ13C values reflect aspects of primate ecologyin addition to diet. In The evolution of Hominin diets (pp. 221–227).Berlin: Springer.

Schoeninger, M. J., Iwaniec, U. T., & Nash, L. T. (1998). Ecological attributesrecorded in stable isotope ratios of arboreal prosimian hair. Oecologia,113(2), 222–230.

Schoeninger, M. J., Moore, J. J., & Sept, J. M. (1999). Subsistence strategiesof two “savanna” chimpanzee populations: The stable isotope evi-dence. American Journal of Primatology, 49, 297–314.

Schoeninger, M. J., Most, C. A., Moore, J. J., & Somerville, A. D. (2016).Environmental variables across Pan troglodytes study sites correspondwith the carbon, but not the nitrogen, stable isotope ratios of chimpan-zee hair. American Journal of Primatology, 78(10), 1055–1069.

Schwarcz, H. P., & Schoeninger, M. J. (1991). Stable isotope analyses inhuman nutritional ecology. American Journal of Physical Anthropology,34(S13), 283–321.

Smith, C. C., Morgan, M. E., & Pilbeam, D. (2010). Isotopic ecology and dietaryprofiles of Liberian chimpanzees. Journal of Human Evolution, 58(1), 43–55.

Sponheimer, M., Loudon, J. E., Codron, D., Howells, M. E., Pruetz, J. D.,Codron, J., … Lee-Thorp, J. A. (2006). Do "savanna" chimpanzees con-sume C4 resources? Journal of Human Evolution, 51(2), 128–133.

Stolwijk, A. M., Straatman, H., & Zielhuis, G. A. (1999). Studying seasonalityby using sine and cosine functions in regression analysis. Journal of Epi-demiology & Community Health, 53(4), 235–238.

Tsutaya, T., Fujimori, Y., Hayashi, M., Yoneda, M., & Miyabe-Nishiwaki, T.(2017). Carbon and nitrogen stable isotopic offsets between diet andhair/feces in captive chimpanzees. Rapid Communications in Mass Spec-trometry, 31(1), 59–67.

Tutin, C. E., Ham, R. M., White, L. J., & Harrison, M. J. (1997). Theprimate community of the Lopé reserve, Gabon: Diets, responsesto fruit scarcity, and effects on biomass. American Journal of Primatology,42(1), 1–24.

van Casteren, A., Oelze, V. M., Angedakin, S., Kalan, A. K., Kambi, M.,Boesch, C., … Kupczik, K. (2018). Food mechanical properties and iso-topic signatures in forest versus savannah dwelling eastern chimpan-zees. Communications Biology, 1(1), 109.

Van der Merwe, N. J., & Medina, E. (1991). The canopy effect, carbon iso-tope ratios and foodwebs in Amazonia. Journal of Archaeological Sci-ence, 18(3), 249–259.

Vogel, E. R., Crowley, B. E., Knott, C. D., Blakely, M. D., Larsen, M. D., &Dominy, N. J. (2011). A noninvasive method for estimating nitrogen bal-ance in free-ranging primates. International Journal of Primatology, 33(3),567–587.

Wallace, M., Jones, G., Charles, M., Fraser, R., Halstead, P., Heaton, T. H., &Bogaard, A. (2013). Stable carbon isotope analysis as a direct means ofinferring crop water status and water management practices. WorldArchaeology, 45(3), 388–409.

Watts, D. P., & Mitani, J. C. (2002). Hunting behavior of chimpanzees atNgogo, Kibale National Park, Uganda. International Journal of Primatol-ogy, 23(1), 1–28.

Wessling, E. G., Deschner, T., Mundry, R., Pruetz, J. D., Wittig, R. M., &Kühl, H. S. (2018). Seasonal variation in physiology challenges thenotion of chimpanzees (Pan troglodytes verus) as a forest-adapted spe-cies. Frontiers in Ecology and Evolution, 6, 60.

Wessling, E. G., Kühl, H. S., Mundry, R., Deschner, T., & Pruetz, J. D.(2018). The costs of living at the edge: Seasonal stress in wildsavanna-dwelling chimpanzees. Journal of Human Evolution,121, 1–11.

Wrangham, R., Conklin, N., Chapman, C., Hunt, K., Milton, K., Rogers, E., …Barton, R. (1991). The significance of fibrous foods for Kibale Forestchimpanzees. Philosophical Transactions of the Royal Society of LondonB: Biological Sciences, 334(1270), 171–178.

Zazzo, A., Harrison, S., Bahar, B., Moloney, A., Monahan, F.,Scrimgeour, C., & Schmidt, O. (2007). Experimental determination ofdietary carbon turnover in bovine hair and hoof. Canadian Journal ofZoology, 85(12), 1239–1248.

SUPPORTING INFORMATION

Additional supporting information may be found online in the Sup-

porting Information section at the end of the article.

How to cite this article: Wessling EG, Oelze VM, Eshuis H,

Pruetz JD, Kühl HS. Stable isotope variation in savanna chim-

panzees (Pan troglodytes verus) indicate avoidance of energetic

challenges through dietary compensation at the limits of the

range. Am J Phys Anthropol. 2019;168:665–675. https://doi.

org/10.1002/ajpa.23782

WESSLING ET AL. 675